import { describe, it, expect, beforeAll } from 'vitest'; import { eq, lt } from 'drizzle-orm'; import { db } from '@/lib/db'; import { gdprExports, aiUsageLedger, user } from '@/lib/db/schema'; import { makePort, makeClient } from '../helpers/factories'; let TEST_USER_ID = ''; beforeAll(async () => { const [u] = await db.select({ id: user.id }).from(user).limit(1); if (!u) throw new Error('No user available; run pnpm db:seed first'); TEST_USER_ID = u.id; }); // Both jobs perform straight DB operations (plus MinIO removeObject for // GDPR). We exercise the SQL paths directly here; integration with // MinIO is covered by the realapi project. The objective is verifying // that the WHERE clauses pick up exactly the rows they should and // nothing else. describe('gdpr-export-cleanup query semantics', () => { it('selects only rows past expiresAt with a storageKey set', async () => { const port = await makePort(); const client = await makeClient({ portId: port.id }); // Three rows: one expired with storageKey (should be picked up), // one expired without storageKey (still building or failed; skip), // one not yet expired (skip). const past = new Date(Date.now() - 60_000); const future = new Date(Date.now() + 60_000); const [expiredWithKey] = await db .insert(gdprExports) .values({ portId: port.id, clientId: client.id, requestedBy: TEST_USER_ID, status: 'ready', storageKey: `${port.slug}/gdpr-exports/${client.id}/test1.zip`, expiresAt: past, }) .returning(); const [expiredNoKey] = await db .insert(gdprExports) .values({ portId: port.id, clientId: client.id, requestedBy: TEST_USER_ID, status: 'failed', storageKey: null, expiresAt: past, }) .returning(); const [stillFresh] = await db .insert(gdprExports) .values({ portId: port.id, clientId: client.id, requestedBy: TEST_USER_ID, status: 'ready', storageKey: `${port.slug}/gdpr-exports/${client.id}/test3.zip`, expiresAt: future, }) .returning(); // The exact predicate the maintenance worker uses: const candidates = await db .select({ id: gdprExports.id }) .from(gdprExports) .where(lt(gdprExports.expiresAt, new Date())); const ids = candidates.map((r) => r.id); expect(ids).toContain(expiredWithKey!.id); expect(ids).toContain(expiredNoKey!.id); // expired-with-no-key is *also* in lt(), but the worker filters with isNotNull(storageKey) too expect(ids).not.toContain(stillFresh!.id); // The full worker filter (expires past, storageKey not null) — only one row. const fullMatch = candidates.filter( (r) => r.id !== expiredNoKey!.id && r.id !== stillFresh!.id, ); expect(fullMatch.map((r) => r.id)).toEqual([expiredWithKey!.id]); }); }); describe('ai-usage-retention query semantics', () => { it('deletes only rows older than the retention window', async () => { const port = await makePort(); // Insert two rows: one fresh, one 100 days old. const fresh = new Date(); const oldDate = new Date(Date.now() - 100 * 24 * 60 * 60 * 1000); const [freshRow] = await db .insert(aiUsageLedger) .values({ portId: port.id, feature: 'ocr', provider: 'openai', model: 'gpt-4o-mini', inputTokens: 100, outputTokens: 50, totalTokens: 150, createdAt: fresh, }) .returning(); const [oldRow] = await db .insert(aiUsageLedger) .values({ portId: port.id, feature: 'ocr', provider: 'openai', model: 'gpt-4o-mini', inputTokens: 100, outputTokens: 50, totalTokens: 150, createdAt: oldDate, }) .returning(); // Mirror the worker's predicate: 90-day cutoff. const cutoff = new Date(Date.now() - 90 * 24 * 60 * 60 * 1000); const deleted = await db .delete(aiUsageLedger) .where(lt(aiUsageLedger.createdAt, cutoff)) .returning({ id: aiUsageLedger.id }); expect(deleted.find((r) => r.id === oldRow!.id)).toBeDefined(); expect(deleted.find((r) => r.id === freshRow!.id)).toBeUndefined(); // The fresh row is still in the table. const survivor = await db .select() .from(aiUsageLedger) .where(eq(aiUsageLedger.id, freshRow!.id)); expect(survivor).toHaveLength(1); }); });